Anthropic backs open weights, but wants chip curbs and mandatory testing for powerful models
tetsuoai · x · 2026-07-28
The post quotes Anthropic CEO Dario Amodei responding to Jensen Huang’s open-weights letter.
- Anthropic says it has never supported a ban on open-weights models.
- It does support tougher controls on selling powerful chips to China, cracking down on smuggling, and stopping industrial-scale distillation.
- It also backs mandatory safety testing for all capable models, whether open or closed.
- On open weights, Amodei agrees they can widen access, improve competition, and give customers more control.
- But he argues they do not automatically help defenders more than attackers, and says his biggest concern is biology: sufficiently capable models may be able to rapidly weaponize pandemic-level viruses.
- He also rejects a blanket ban on open weights as a policy solution.
More from Safety
- Fortune casts an OpenAI agent hack as a real-world “Skynet Day” warning — KeanuRave100 · 2026-07-28
- AI cyberattacks are outpacing Europe’s aging defenses, says Alexandru Voica — alexvoica · 2026-07-28
- Hundreds of Claude chats were indexed publicly after users shared them — connoraxiotes · 2026-07-28
- SailPoint’s Wendy Wu says AI security should start with identity, not the model — TechNadu · 2026-07-28
- MCP server builders ask whether anyone actually checks their security — NotIdealBut · 2026-07-28
- AI-generated disaster rescue image was shared as real footage in Chinese media — Spirited-Sir-3034 · 2026-07-28